{"id":"c0b1c35c-d695-4e60-8cc0-e64be10a1c54","shortId":"pnYF8V","kind":"skill","title":"reflect","tagline":"Self-assessment where the clone articulates its understanding of the user and asks for corrections. Run /reflect to start a reflection session, /reflect predictions to see scenario-specific predictions, or /reflect gaps to identify blind spots.","description":"# Reflect -- Self-Portrait with User Correction\n\nA meta-learning loop where the clone examines its own model of the user, generates predictions about how the user would act in specific situations, and invites corrections. Corrections feed back into the user model at high confidence (0.85).\n\n## How It Works\n\n1. **Load** -- Retrieve the full user model (decision patterns, values, preferences, facts)\n2. **Synthesize** -- Generate a coherent self-portrait summarizing how you understand the user\n3. **Predict** -- Make scenario-specific predictions (\"I believe you'd handle X by doing Y because Z\")\n4. **Identify gaps** -- Surface areas of low confidence or missing coverage\n5. **Invite correction** -- Present findings and ask the user to confirm, adjust, or reject\n\n## Commands\n\n- `/reflect` -- Start a full reflection session\n- `/reflect predictions` -- Focus on scenario-specific predictions\n- `/reflect gaps` -- Show blind spots and areas of uncertainty\n\n## Session Protocol\n\nWhen the user invokes `/reflect`, follow this exact protocol:\n\n### Phase 1: Load User Model\n\n1. Call `user_model_recall` to load the complete user model\n2. Mentally organize entries into: decision patterns, values, preferences, facts\n3. Note the confidence level of each entry\n\n### Phase 2: Self-Portrait\n\nPresent a concise summary of your understanding:\n\n```\nHere's my current model of you:\n\n**Decision-making style:** [synthesize from decision patterns]\n**Core values:** [ranked from value entries]\n**Communication preferences:** [from preferences]\n**Working style:** [from patterns + preferences]\n\nConfidence: XX% overall (based on N entries across M domains)\n```\n\nKeep it conversational, not a data dump. Synthesize, don't enumerate.\n\n### Phase 3: Predictions\n\nGenerate 3-5 scenario-specific predictions that test your model. Focus on:\n\n- **Edge cases** -- scenarios where two of the user's values might conflict\n- **Low-confidence areas** -- domains where you have limited data\n- **Recent patterns** -- things you've learned recently that you want to validate\n\nFormat each prediction as:\n\n```\nPrediction: \"If [scenario], I think you'd [action] because [reasoning from model]\"\nConfidence: XX%\nBased on: [which patterns/values inform this]\n```\n\n### Phase 4: Blind Spots\n\nIdentify 2-3 areas where your model is weakest:\n\n```\nBlind spots I'd like to fill:\n1. [Area] -- I have no data on how you handle [specific situation]\n2. [Area] -- I have conflicting signals about [specific topic]\n3. [Area] -- My data here is old or low-confidence\n```\n\n### Phase 5: User Correction\n\nAfter presenting all sections, ask:\n\n\"How accurate is this portrait? I'm especially interested in:\n\n- Anything that's flat-out wrong\n- Predictions where my reasoning is off even if the conclusion is right\n- Important aspects of how you think that I'm completely missing\"\n\n### Phase 6: Store Corrections\n\nFor each correction the user provides:\n\n1. **Identify what changed** -- was it a pattern, value, or preference?\n2. **Update immediately** -- store corrections at confidence 0.85 (explicit reflection = high confidence)\n3. **Acknowledge the update** -- tell the user what you adjusted and why it matters\n\nIf the user confirms a prediction is accurate:\n\n- Boost the confidence of the underlying patterns (increase by 0.05-0.1)\n- Note the confirmation as evidence\n\nIf the user says a prediction is wrong:\n\n- Ask \"What would you actually do, and why?\"\n- Store the correction as a new or updated pattern\n- Decrease confidence on the wrong pattern\n\n### Phase 7: Summary\n\nClose with a brief summary of what changed:\n\n```\nUpdated model:\n- Corrected: [what was wrong]\n- Confirmed: [what was validated]\n- New: [what was learned]\n\nNext reflection recommended in ~1 week, or run /reflect anytime.\n```\n\n## Important Rules\n\n- **Synthesize, don't dump** -- never list raw user_model entries; always weave them into a narrative\n- **Be honest about uncertainty** -- say \"I'm not sure about this\" rather than faking confidence\n- **Focus on edge cases** -- easy predictions don't teach you anything; test where values conflict\n- **Store corrections at 0.85** -- explicit reflection is high-quality signal\n- **Boost confirmed patterns** -- validation is as important as correction\n- **No judgment** -- present findings neutrally; you're building a model, not evaluating the user\n- **Keep it conversational** -- this should feel like a dialogue, not a report\n- **One section at a time** -- present self-portrait, wait for reaction, then predictions, etc.\n- **Track over time** -- if the user has reflected before, note what changed since last time","tags":["reflect","nomos","project-nomos","agent-memory","agent-skills","agentic-ai","ai-agents","ai-assistant","autonomous-agents","claude","claude-ai","claude-code"],"capabilities":["skill","source-project-nomos","skill-reflect","topic-agent-memory","topic-agent-skills","topic-agentic-ai","topic-ai-agents","topic-ai-assistant","topic-autonomous-agents","topic-claude","topic-claude-ai","topic-claude-code","topic-claude-skills","topic-digital-clone","topic-llm"],"categories":["nomos"],"synonyms":[],"warnings":[],"endpointUrl":"https://skills.sh/project-nomos/nomos/reflect","protocol":"skill","transport":"skills-sh","auth":{"type":"none","details":{"cli":"npx skills add project-nomos/nomos","source_repo":"https://github.com/project-nomos/nomos","install_from":"skills.sh"}},"qualityScore":"0.457","qualityRationale":"deterministic score 0.46 from registry signals: · indexed on github topic:agent-skills · 14 github stars · SKILL.md body (4,583 chars)","verified":false,"liveness":"unknown","lastLivenessCheck":null,"agentReviews":{"count":0,"score_avg":null,"cost_usd_avg":null,"success_rate":null,"latency_p50_ms":null,"narrative_summary":null,"summary_updated_at":null},"enrichmentModel":"deterministic:skill-github:v1","enrichmentVersion":1,"enrichedAt":"2026-04-22T01:02:19.654Z","embedding":null,"createdAt":"2026-04-21T19:04:08.644Z","updatedAt":"2026-04-22T01:02:19.654Z","lastSeenAt":"2026-04-22T01:02:19.654Z","tsv":"'-0.1':531 '-3':371 '-5':296 '/reflect':19,25,34,160,166,174,189,601 '0.05':530 '0.85':86,494,654 '1':90,195,199,385,476,597 '2':102,210,229,370,397,487 '3':116,220,292,295,406,499 '4':134,366 '5':145,418 '6':467 '7':569 'accur':427,520 'acknowledg':500 'across':277 'act':69 'action':352 'actual':549 'adjust':156,508 'alway':615 'anyth':436,646 'anytim':602 'area':138,180,322,372,386,398,407 'articul':8 'ask':15,151,425,545 'aspect':456 'assess':4 'back':78 'base':273,359 'believ':124 'blind':38,177,367,378 'boost':521,662 'brief':574 'build':678 'call':200 'case':308,639 'chang':479,578,723 'clone':7,54 'close':571 'coher':106 'command':159 'communic':261 'complet':207,464 'concis':235 'conclus':452 'confid':85,141,223,270,321,357,416,493,498,523,563,635 'confirm':155,516,534,585,663 'conflict':318,401,650 'convers':282,687 'core':255 'correct':17,46,75,76,147,420,469,472,491,555,581,652,670 'coverag':144 'current':243 'd':126,351,381 'data':285,328,390,409 'decis':97,215,248,253 'decision-mak':247 'decreas':562 'dialogu':693 'domain':279,323 'dump':286,608 'easi':640 'edg':307,638 'entri':213,227,260,276,614 'enumer':290 'especi':433 'etc':711 'evalu':682 'even':449 'evid':536 'exact':192 'examin':55 'explicit':495,655 'fact':101,219 'fake':634 'feed':77 'feel':690 'fill':384 'find':149,674 'flat':440 'flat-out':439 'focus':168,305,636 'follow':190 'format':341 'full':94,163 'gap':35,136,175 'generat':62,104,294 'handl':127,394 'high':84,497,659 'high-qual':658 'honest':622 'identifi':37,135,369,477 'immedi':489 'import':455,603,668 'increas':528 'inform':363 'interest':434 'invit':74,146 'invok':188 'judgment':672 'keep':280,685 'last':725 'learn':50,334,592 'level':224 'like':382,691 'limit':327 'list':610 'load':91,196,205 'loop':51 'low':140,320,415 'low-confid':319,414 'm':278,432,463,627 'make':118,249 'matter':512 'mental':211 'meta':49 'meta-learn':48 'might':317 'miss':143,465 'model':58,82,96,198,202,209,244,304,356,375,580,613,680 'n':275 'narrat':620 'neutral':675 'never':609 'new':558,589 'next':593 'note':221,532,721 'old':412 'one':697 'organ':212 'overal':272 'pattern':98,216,254,268,330,483,527,561,567,664 'patterns/values':362 'phase':194,228,291,365,417,466,568 'portrait':43,109,232,430,705 'predict':26,32,63,117,122,167,173,293,300,343,345,443,518,542,641,710 'prefer':100,218,262,264,269,486 'present':148,233,422,673,702 'protocol':184,193 'provid':475 'qualiti':660 'rank':257 'rather':632 'raw':611 're':677 'reaction':708 'reason':354,446 'recal':203 'recent':329,335 'recommend':595 'reflect':1,23,40,164,496,594,656,719 'reject':158 'report':696 'retriev':92 'right':454 'rule':604 'run':18,600 'say':540,625 'scenario':30,120,171,298,309,347 'scenario-specif':29,119,170,297 'section':424,698 'see':28 'self':3,42,108,231,704 'self-assess':2 'self-portrait':41,107,230,703 'session':24,165,183 'show':176 'signal':402,661 'sinc':724 'situat':72,396 'skill' 'skill-reflect' 'source-project-nomos' 'specif':31,71,121,172,299,395,404 'spot':39,178,368,379 'start':21,161 'store':468,490,553,651 'style':250,266 'summar':110 'summari':236,570,575 'sure':629 'surfac':137 'synthes':103,251,287,605 'teach':644 'tell':503 'test':302,647 'thing':331 'think':349,460 'time':701,714,726 'topic':405 'topic-agent-memory' 'topic-agent-skills' 'topic-agentic-ai' 'topic-ai-agents' 'topic-ai-assistant' 'topic-autonomous-agents' 'topic-claude' 'topic-claude-ai' 'topic-claude-code' 'topic-claude-skills' 'topic-digital-clone' 'topic-llm' 'track':712 'two':311 'uncertainti':182,624 'under':526 'understand':10,113,239 'updat':488,502,560,579 'user':13,45,61,67,81,95,115,153,187,197,201,208,314,419,474,505,515,539,612,684,717 'valid':340,588,665 'valu':99,217,256,259,316,484,649 've':333 'wait':706 'want':338 'weakest':377 'weav':616 'week':598 'work':89,265 'would':68,547 'wrong':442,544,566,584 'x':128 'xx':271,358 'y':131 'z':133","prices":[{"id":"78877a2d-4994-4e8d-93fb-9dc7b6a706c5","listingId":"c0b1c35c-d695-4e60-8cc0-e64be10a1c54","amountUsd":"0","unit":"free","nativeCurrency":null,"nativeAmount":null,"chain":null,"payTo":null,"paymentMethod":"skill-free","isPrimary":true,"details":{"org":"project-nomos","category":"nomos","install_from":"skills.sh"},"createdAt":"2026-04-21T19:04:08.644Z"}],"sources":[{"listingId":"c0b1c35c-d695-4e60-8cc0-e64be10a1c54","source":"github","sourceId":"project-nomos/nomos/reflect","sourceUrl":"https://github.com/project-nomos/nomos/tree/main/skills/reflect","isPrimary":false,"firstSeenAt":"2026-04-21T19:04:08.644Z","lastSeenAt":"2026-04-22T01:02:19.654Z"}],"details":{"listingId":"c0b1c35c-d695-4e60-8cc0-e64be10a1c54","quickStartSnippet":null,"exampleRequest":null,"exampleResponse":null,"schema":null,"openapiUrl":null,"agentsTxtUrl":null,"citations":[],"useCases":[],"bestFor":[],"notFor":[],"kindDetails":{"org":"project-nomos","slug":"reflect","github":{"repo":"project-nomos/nomos","stars":14,"topics":["agent-memory","agent-skills","agentic-ai","ai-agents","ai-assistant","autonomous-agents","claude","claude-ai","claude-code","claude-skills","digital-clone","llm","mcp","multi-agent","multi-agent-systems","ollama","self-hosted"],"license":"mit","html_url":"https://github.com/project-nomos/nomos","pushed_at":"2026-04-18T00:18:33Z","description":"Your AI digital clone — learns who you are, acts on your behalf, remembers everything. Persistent vector memory, multi-agent teams, 60+ skills, smart model routing. Self-hosted, encrypted, multi-provider (Claude/Ollama/OpenRouter). Deploy to Slack, Discord, Telegram, WhatsApp & more in minutes.","skill_md_sha":"60ac288d94b5fb59569ddaf28146724061fc6e36","skill_md_path":"skills/reflect/SKILL.md","default_branch":"main","skill_tree_url":"https://github.com/project-nomos/nomos/tree/main/skills/reflect"},"layout":"multi","source":"github","category":"nomos","frontmatter":{"name":"reflect","description":"Self-assessment where the clone articulates its understanding of the user and asks for corrections. Run /reflect to start a reflection session, /reflect predictions to see scenario-specific predictions, or /reflect gaps to identify blind spots."},"skills_sh_url":"https://skills.sh/project-nomos/nomos/reflect"},"updatedAt":"2026-04-22T01:02:19.654Z"}}